AI-Powered Sketching Tool Inkspire Breaks Design Fixation and Boosts Creativity

Traditional text-to-image AI tools often trap designers in a loop of minor refinements, but Inkspire’s sketch-driven workflow transforms abstract ideas into creative breakthroughs, fostering an iterative and fluid design process.

The Inkspire interface. The designer may use the Analogical Panel (a) to ideate analogical inspirations for abstract concepts (e.g., “protective car”→“tortoise car”). The designer may sketch on the Sketching Panel (b) to iteratively guide AI design generations. For each iteration, we display a sketch scaffolding under the canvas. This scaffolding is created through abstracting AI designs into lower fidelity. Finally, the designer may view the history of iterations on the Evolution Panel (c).

*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as definitive, used to guide development decisions, or treated as established information in the field of artificial intelligence research.

With rapid advancements in text-to-image (T2I) models, designers now have unprecedented tools to translate abstract concepts into detailed visual representations. While these models enhance productivity and offer creative inspiration, they also introduce challenges such as design fixation and the unnatural reliance on text-based prompting. Designers often struggle to explore beyond their initial concepts, making only incremental refinements rather than expanding their creative horizons.

Research presented in a paper posted to the arXiv preprint* server addresses these challenges by introducing Inkspire, a sketch-driven tool that integrates analogical inspirations and a complete sketch-to-design-to-sketch feedback loop. Conducted by researchers from Carnegie Mellon University and the Toyota Research Institute, this study explores how designers interact with generative AI and evaluates the effectiveness of Inkspire in fostering creativity and overcoming design fixation. The study found that designers using Inkspire demonstrated significantly greater exploration and inspiration scores compared to those using conventional ControlNet-based T2I tools.

Challenges in Generative AI for Design

Generative AI tools offer powerful visual generation capabilities, yet their interaction paradigms often limit creative workflows. Designers report that using text prompts to generate images feels unnatural compared to sketching, which is their traditional mode of ideation. Moreover, T2I models struggle with abstract concepts; for instance, asking AI to generate a "protective" car often results in uninspired or generic outputs. Lastly, AI-generated images frequently appear too polished or "complete," making it difficult for designers to iterate upon them. These issues contribute to design fixation, where users get stuck refining a narrow set of ideas rather than exploring a broad design space. Inkspire specifically addresses this by using an analogy-driven approach, helping designers explore novel inspirations beyond their initial thoughts.

Inkspire: A Sketch-Driven Approach

Inkspire was developed to provide an alternative method for interacting with AI in a more fluid and natural way. Instead of text-based prompting, designers can generate analogical inspirations—concepts drawn from unrelated domains, such as nature, architecture, or fashion—that serve as seeds for new ideas. By leveraging Large Language Models (LLMs), Inkspire identifies visually concrete objects related to an abstract theme, guiding designers toward novel interpretations. For example, the concept of "protectiveness" might be translated into inspirations such as "tortoise," "bunker," or "armor." Designers can then explore variations of these inspirations, using a Sketch2Design interface to iteratively refine their ideas. Unlike traditional text-prompting methods, users of Inkspire engage in an iterative process, making small modifications and receiving AI-generated suggestions at each step, fostering a more dynamic workflow.

Example designs created by participants using Inkspire for the design tasks of designing a fluid chair and a serene lamp. The final participant-generated sketch is shown on the top, and the generated T2I image is shown on the bottom, along with the selected analogy word chosen by the participant.

Sketch2Design and Design2Sketch Feedback Loop

Inkspire introduces a bidirectional workflow, where sketches can be transformed into AI-generated images and AI-generated images can be abstracted back into sketches. This Sketch2Design and Design2Sketch cycle allows designers to maintain a fluid iteration process. In Sketch2Design, users sketch a rough outline, and the AI generates refined concept variations based on the selected analogies. Inkspire enhances this process by dynamically adjusting the AI’s sensitivity to sketches using a ControlNet-based model, ensuring that rough initial strokes lead to progressively more refined results. In Design2Sketch, the AI-generated images are converted into low-fidelity sketches, helping designers avoid fixation on photorealistic renderings and instead focus on conceptual exploration. This is achieved through a novel approach combining semantic segmentation and edge extraction, ensuring that key structural features remain visible in the abstracted sketch while unnecessary details are removed. This iterative approach ensures that designers retain creative control while benefiting from AI-generated suggestions.

User Study and Findings

To evaluate Inkspire’s effectiveness, the researchers conducted a within-subjects study comparing Inkspire with a conventional ControlNet-based T2I system. Twelve participants, including both professional designers and novices, completed design tasks using both tools. Measures included creativity support, human-AI collaboration, final design quality, and user experience satisfaction.

The results showed that Inkspire significantly improved designers’ ability to explore novel ideas and engage in more iterative design processes. Participants using Inkspire demonstrated statistically significant increases in exploration (p < 0.01) and inspiration (p < 0.01) compared to the baseline system, as measured by the Creativity Support Index. Participants using Inkspire reported higher inspiration and exploration scores than those using ControlNet. Inkspire also enhanced co-creation with AI, with users feeling a greater sense of partnership, controllability, and ownership over their designs. Additionally, participants using Inkspire engaged in more iterative sketching, making fewer but more deliberate strokes while allowing the AI to refine and evolve their designs dynamically. The study found that designers generated a larger number of concept sketches with Inkspire, indicating a more diverse and exploratory workflow compared to the baseline system.

Addressing Design Fixation

One of Inkspire’s key contributions is its ability to mitigate design fixation. Traditional T2I tools encourage users to tweak prompts incrementally, which often results in minor variations of an initial idea rather than broad conceptual exploration. Inkspire’s analogy-driven approach helps designers break free from fixed thought patterns by presenting unexpected but relevant inspirations. Additionally, the low-fidelity sketching scaffold prevents users from becoming too fixated on polished AI-generated images, allowing for greater flexibility and continued iteration. User interactions revealed that those using ControlNet tended to sketch full designs before generating images. In contrast, Inkspire users followed a more fluid, iterative process, refining their ideas with each AI-assisted step.

Future Directions and Implications

The findings suggest that multimodal AI interfaces can better support creative workflows than text-only interaction models. Future work on Inkspire could enhance its interaction techniques, expand analogy sources, and refine AI-driven feedback mechanisms. Potential improvements include allowing users to explore multiple analogical inspiration branches in parallel and introducing finer-grained sketch controls for specifying materials or mechanical constraints. The research also highlights the broader implications for co-creative AI systems, suggesting that AI should serve as an adaptive collaborator rather than a static generator. By prioritizing iterative and sketch-driven interactions, AI-powered tools can better integrate into designers’ creative processes, leading to richer and more innovative design outcomes.

Conclusion

Inkspire represents a significant step toward more intuitive and exploratory generative AI tools for design. By incorporating analogy-driven inspiration and sketch-based interactions, it enables designers to engage in a more iterative, flexible, and creative workflow. The research demonstrates that sketch-driven AI tools can help designers generate a broader range of ideas, overcome fixation, and interact with AI in a more natural and meaningful way. With its ability to enhance inspiration, exploration, and human-AI collaboration, Inkspire sets a precedent for the development of future co-creative AI tools. As generative AI continues to evolve, approaches like Inkspire pave the way for more effective human-AI collaboration in design and other creative fields.

*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as definitive, used to guide development decisions, or treated as established information in the field of artificial intelligence research.